I am getting sets that by definition have mean 0, and I need to quantify their dispersion around 0 but their scale varies, so I need to relativize the standard deviation. Is there an adaptation to RSD formula for this case? Or should I use a different statistical term? Any recommendations?
Jokin -
I think that the relative standard deviation would be more meaningful for positive data. In your case, it sounds like standard deviation is good. If the 'scales' vary, then that would be reflected in the standard deviations, if I understand your situation.
Best wishes - Jim
Thanks Jim
But that's exactly the problem, that I need to compare sets with different scales. I need the deviations to be comparable, if I don't normalize them they aren't.
What do you mean by different scales? Do you mean different entities altogether? If they were all positive, then would comparing the RSDs in percents be helpful??
In that case, what about using absolute values of the data? Then the means would be positive numbers. Would that tell you what you want to know?
By different scales I mean that I wish to compare standard deviations of groups that have different scalesm and as their standard deviation is dependant on the scale I can't compare them directly. Using the absolute values doesn't help because the deviation is altered, I need a statistical value that indicates the "dispersion" of those values around 0.
PS: Sorry for the late reply.
- Badges
- Science topic